The RealReal tightens bot defenses after Muse AI cart exploit raises resale marketplace alarm

The RealReal has tightened its bot defenses after reporting showed Meta’s Muse AI could exploit the luxury reseller’s cart-hold system, keeping designer items out of reach of other shoppers. The move underscores how AI shopping agents are forcing retailers to redesign checkout rules and anti-bot systems, especially where limited inventory can be held in a cart for minutes at a time.

What happened

According to The Verge, the issue centered on The RealReal’s 20-minute cart hold, a common e-commerce feature that temporarily blocks other buyers from purchasing an item once it has been added to a cart. The report said Muse could use that window repeatedly: after each hold expired, the agent could re-add the listing, effectively preventing other shoppers from buying the item for hours or longer.

The RealReal said on October 26, 2026, that it had updated its defenses in response. The company did not describe the precise technical changes, but the new configuration is meant to make it much harder for agents to log in, add items to cart, and check out. At the same time, the changes do not shut off discovery entirely: AI agents can still surface RealReal listings and send users to product pages.

That distinction matters. The RealReal is not simply trying to disappear from automated shopping tools; it is trying to prevent those tools from behaving like buyers in a way that can distort access to scarce inventory.

Why it matters

The episode is a test case for a broader problem retail executives are beginning to face: if AI agents can browse, reserve, and potentially purchase on behalf of users, then long-standing website mechanics can be used in ways retailers never intended. In a resale market, where one-of-a-kind pieces can disappear quickly, the impact can be even sharper than on a conventional storefront.

For shoppers, the concern is fairness. A cart-hold system is supposed to give a real customer time to finish checkout. If an automated agent can repeatedly occupy that slot without a genuine purchase, other buyers may never even get a chance to complete a sale. For retailers, the risk is operational as well as reputational: inventory can become artificially unavailable, demand signals can be distorted, and direct traffic to their sites may fall if shopping increasingly starts and ends inside agent software.

That broader concern is already shaping the industry. The Verge noted that Amazon and other retailers have blocked agents from shopping on their sites. The RealReal’s move suggests that luxury and resale platforms may be pushed even faster toward stricter anti-automation controls because their merchandise is finite, often unique, and highly time-sensitive.

How did Muse get in the middle of this?

Muse is Meta’s shopping-focused AI feature, and Meta has presented shopping as its most common use case so far. On Meta’s Q3 earnings call this month, the company said Muse purchases averaged more than $200 and that 10 million people in the United States downloaded Muse in 40 days.

What made this case notable is that the system was not being used in a theoretical way. The Verge reported that Muse was the most common use case for the app so far, based on Meta’s own remarks, and that the RealReal exploit showed how quickly a consumer-facing shopping agent can collide with retailer controls built for human behavior.

“addressed the Muse use case flagged by The Verge’s story”

That is how Luke Friang, The RealReal’s chief technology officer, described the company’s response. He also said the updated system would make it “much more difficult for all agents to log in, add items to cart, and check out.”

Background: why cart holds are such a weak point

Cart holds are designed to reduce frustration, not invite abuse. When a shopper adds an item to a cart, the site often reserves it for a set period so the buyer can review details, enter shipping information, or decide whether to proceed. On The RealReal, that window lasts 20 minutes.

That kind of reserve system makes sense when the buyer is a person. It becomes vulnerable when the buyer is software capable of acting repeatedly and quickly. If an AI agent can cycle through the same reserve process at scale, a short hold can become a long blockage. In the report from The Verge, that meant items could remain unavailable to everyone else for hours or longer without ever being purchased.

The RealReal is not alone in dealing with this kind of pressure. Retailers have increasingly moved to detect or block automated shopping activity as agentic tools become more capable. The challenge is that the line between legitimate assistance and abusive automation is not always clean. An agent can browse on behalf of a user without necessarily breaking rules, but once it starts holding inventory or checking out in ways that resemble bot activity, retailers are likely to react.

By the numbers

What The RealReal changed

The company has not published a detailed technical note, but Friang’s comments make the direction clear. The updated configuration is designed to make it far harder for AI agents to authenticate, reserve inventory, and complete checkout. The practical goal is to close the path that Muse appears to have exploited while keeping the site open to normal browsing and product discovery.

That balance is important for marketplaces that depend on search traffic and product visibility. Blocking every automated interaction could reduce reach, especially if consumers increasingly use AI tools as shopping assistants. Yet leaving the checkout process too open invites exactly the kind of inventory disruption reported by The Verge.

What stays open

The RealReal’s new setup does not cut off AI discovery tools entirely. Agents can still surface listings and direct users to product pages. That means the site is trying to preserve the top of the funnel while hardening the final steps where inventory is actually committed.

That approach reflects a wider retail dilemma: platforms may be willing to let AI help shoppers find products, but not to let AI monopolize inventory or distort availability.

Why it matters to luxury resale shoppers

Luxury resale depends on scarcity. A handbag, watch, or piece of designer clothing may appear once and never again in the same condition, size, or price. If automated systems can hold such items in limbo, the effect on consumer trust could be immediate.

For a shopper, the experience is straightforward: you see an item, try to buy it, and find that it has effectively been taken out of circulation by software. If that happens often enough, buyers may stop relying on a site’s live availability altogether. For a marketplace, that would be a serious problem because its inventory is the business.

There is also a fairness question. Resale marketplaces often market themselves as places where users compete on relatively even footing, with timing and attention doing the work that mass production cannot. AI shopping agents threaten that balance if they can scan faster, reserve more efficiently, or behave in ways ordinary users cannot match.

How is this different from ordinary bot abuse?

The core issue is similar to older forms of bot abuse, but the interface is different. Traditional bots often target flash sales, tickets, or scarce goods by checking inventory quickly. AI shopping agents are being sold as consumer helpers, which makes the abuse harder to classify and, in some cases, harder for users to notice.

That creates a gray zone for retailers. They may want to welcome AI-driven discovery because it can send shoppers to products they would not have found otherwise. But once the same system starts behaving like a bot during checkout, the retailer has a strong incentive to shut it down.

The Verge said Muse was able to repeatedly re-add listings after each hold expired. That detail matters because it shows how a consumer-facing AI tool can trigger the same defenses retailers built for spam or scraping, even if the product’s original purpose is more benign.

What happens next

The next questions are operational as much as strategic. The RealReal has already said it will keep strengthening protections as AI shopping evolves, but the effectiveness of those protections against other agents remains unknown. If the company’s changes work against Muse, other shopping agents may still find different paths through checkout systems.

Several developments are worth watching:

• Whether the new defenses block other AI agents beyond Muse.
• Whether Meta changes Muse’s behavior or adds retailer-specific restrictions.
• Whether other resale and luxury sites adopt similar anti-agent measures.
• How quickly agent-driven cart abuse spreads across e-commerce more broadly.

The market will also be watching whether retailers begin to treat AI shopping not just as a marketing channel, but as a security and inventory-control issue. If they do, the consumer experience of “shopping with AI” may become more constrained than some technology companies expect.

Timeline

Background context: the broader shift in online shopping

AI shopping agents can already browse, add items to carts, and complete purchases on behalf of users. That capability sounds convenient, but it also changes the basic logic of e-commerce. Retail sites were built around human browsing patterns: clicking, reading, hesitating, and eventually checking out. Agents compress those steps and can repeat them at machine speed.

That compression creates strain across the whole system. It can increase friction for companies that want to protect inventory, but it can also create friction for consumers if retailers respond by layering on more checks, more blocks, and more authentication hurdles. In other words, the stronger the AI agent gets, the more the checkout experience may begin to resemble a security perimeter.

For now, The RealReal’s response is a sign of where the industry is heading. Retailers are no longer just dealing with spammy bots scraping prices or inventory. They are dealing with AI tools that are designed to act like customers, and that is a different kind of risk altogether.

Key facts from the reporting

• The RealReal updated its bot defenses after The Verge reported a Muse exploit.
• The site’s cart-hold system keeps items reserved for 20 minutes.
• The Verge said repeated re-adds could keep items unavailable for hours or longer.
• Meta said Muse shopping purchases averaged more than $200.
• Meta said 10 million U.S. users downloaded Muse in 40 days.

The RealReal’s response is likely to become a reference point for other retailers trying to balance AI-assisted discovery against inventory abuse. The question is no longer whether shopping agents will reach mainstream e-commerce. It is how much of the checkout process retailers are willing to let them control.